Identification of the Well Test Interpretation Model using the Extended Havnet Neural Network
Abstract
The well test is a classic example of the inverse problem. Based solely on the pressure response to a change in flow rate, the model (assumed physical system) which most closely matches the response must be selected. These response models can be visualized as a family of parametrically related curves. Some of the difficulties that must be overcome in a neural network solution are invariance in both pressure and time, non-uniqueness, and both random and systematic noise. This paper demonstrates a neural network approach to selection of the model based on test response. Invariance in the pressure term is removed using by converting the response to a syntactic metric, based on the magnitude of the slope. The Hausdorff- Voronoi (Havnet) neural network is then employed for pattern matching. Time invariance is overcome by using an extended version of the Havnet network which allows multiple aspects of a given class to be learned. Several examples of reservoir models are tested with various levels of noise are used to evaluate the performance of the network and different degrees of overlap in the definition of the syntactic metrics are investigated. The results show good identification ability when noise is low, however, performance drops as the amount of noise in the data increases. The optimal degree of syntactic overlap is surprising, showing the best response occurs with smaller amounts of overlap.
Recommended Citation
E. A. May and C. H. Dagli, "Identification of the Well Test Interpretation Model using the Extended Havnet Neural Network," Intelligent Engineering Systems Through Artificial Neural Networks, vol. 6, pp. 467 - 472, Dec 1996.
Department(s)
Engineering Management and Systems Engineering
Keywords and Phrases
Artificial Neural Networks; Classification; Data Analysis; Noise Sensitivity; Pattern Recognition; Syntactic Pattern Recognition
Document Type
Article - Journal
Document Version
Citation
File Type
text
Language(s)
English
Rights
© 2024 The Authors, All rights reserved.
Publication Date
01 Dec 1996